1. Notable to my mind are Cummins (1989), who addresses the use of the word 'representation' in cognitive science in ways that are similar to those suggested here, and yon Eckardt (1993), who takes a more comprehensive look at the philosophy of cognitive science as philosophy of science. While I think that yon Eckardt is inadequately sensitive to subtle problems in the notion of representation, this kind of book is a very welcome development in the field. [BACK]

1. Fodor (1987: 6) says that "the predictive adequacy of commonsense psychology is beyond rational dispute." [BACK]

2. Pylyshyn (1984: 7), for example, writes that there are "regularities and generalizations which can be captured using cognitive terms that could not be captured in descriptions using behavioral or physical (neurophysiological) terms." And Fodor (1987: 9) writes in a similar vein that "we have no idea of how to explain ourselves to ourselves except in a vocabulary which is saturated with belief/desire psychology." [BACK]

3. Fodor (1987: 15) thus writes that "a lot of what common sense believes about the attitudes must surely be false," and Pylyshyn (1980: 113) opines that it may be that folk psychology's "set of terms needs to be augmented or pruned, that many of the beliefs expressed in folk psychology are either false or empirically empty, and that many of its explanations are either incomplete or circular." [BACK]

4. Fodor (1981: 18-19), for example, writes that there are "lots of interesting generalizations to state about how the propositional attitudes that an organism has are affected by aspects of its experience, and its genetic endowment, continue

and by the other propositional attitudes that it entertains. One of the goals of a (cognitive) psychology is to state the generalizations and, whenever possible, to systematize and explain them." [BACK]

5. Pylyshyn (1980:112) writes that "its most serious shortcoming, from the point of view of the scientific enterprise, is that the collection of loose generalizations that makes up this informal body of knowledge is not tied together into an explicit system." [BACK]

6. Fodor's "cruder but more intelligible" gloss would also seem to indicate that beliefs with different contents are typified by the same functional relationship R , but differ with respect to the type of representation to which the organism is related. So if Jones's believing that two is a prime number consists in Jones standing in relation R to a representation of a type MP that means that two is a prime number, then Jones's believing that it is raining must consist in Jones being in the same relation R to a representation of a different type, MP ^* , such that tokens of MP ^* mean that it is raining. Conversely, when Jones has different cognitive states that share a common content, the different states are characterized by different relations to tokens of the same symbol type. For example, if Jones both believes and hopes that it is raining, his believing that it is raining consists in his being in relation R to a token of type MP ^* , and his hoping that it is raining consists in his being in some other relation R ^* to a token of type MP ^* . [BACK]

7. First, it is not clear whether the same relation R is characteristic of belief for different cognizers, or whether the functional relations characteristic of propositional attitudes such as believing or desiring may differ from organism to organism. That is, it is unclear whether Smith's believing that two is a prime number (as opposed to Jones's believing it) would involve Smith being in the same relation R to a mental representation meaning that two is a prime number that Jones is in when he believes that two is a prime number, or whether it would involve Smith being in some other relation S to such a representation. This question is highly significant in its bearing upon how CTM might guide empirical research, but will not have an effect upon the results of this inquiry.

Fodor's articulation of the nature of cognitive states is also unclear on the question of whether bearing a particular functional relation to a mental representation is to count as a sufficient condition for being in a particular cognitive state, or whether it is merely a necessary condition. What Fodor (1987: 17) says is that " to believe that such and such is to have a mental symbol that means that such and such tokened in your head in a certain way." But it is not fully clear whether Fodor means to say ( a ) "to believe that such and such is just to have a mental symbol that means that such and such tokened in your head in a certain way," or ( b ) "to believe that such and such is to be in a state that includes having a mental symbol that means such and such tokened in your head in a certain way." On interpretation (a), Fodor's account offers a sufficient condition for an organism O's being in a particular cognitive state; on interpretation (b), it offers only a necessary condition. Interpretation ( a ) seems the more natural reading of the text in question, but the reader is naturally guarded about drawing too strong an inference from a characterization that is described by the author as "crude." break [BACK]

8. Fodor (1981: 318, n. 2) reserves the spelling of 'intentionality' with a t for "contexts connected with intent." [BACK]

9. There is, of course, a significant history of controversy about whether intentional explanation is a form of causal explanation, and the issues that stand between, say, Vienna Circle positivism on the one hand and Ryle, Wittgenstein, and the Verstehen tradition on the other. In particular, there has been controversy about the nature of intentional explanation, notably about whether such explanations provide only reasons for action, only causes of action, or both. Much of the continental tradition in philosophy and the human sciences has held that explanation in the human sciences gives reasons for action and provides interpretive understanding (Verstehen), whereas explanation in the natural sciences gives causes of events and provides explanations. In the English-speaking world, Wittgenstein, Ryle, and other ordinary-language philosophers have advanced similar views to the effect that intentional explanations provide reasons for action and do not name causes. Much of the English-speaking philosophical tradition, on the other hand, has followed the logical positivists in assuming that the causal-nomological mode of explanation employed in the natural sciences is normative for the application of the word 'science', with the implication that psychological explanation must either be causal and nomological or else be unscientific. (The contribution of CTM to this discussion will make up the greater part of chapter 2 of this book.) [BACK]

10. While computational procedures (such as integration techniques and column addition) and formalized mathematical systems (like Hilbert's geometry) may both be said to involve "formal rules," the "rules" they involve are of different sorts. Formalized mathematical systems such as formal logic and formal geometry contain derivation-licensing rules, but no rules of procedure . That is, they involve rules that govern the validity of derivations, but give no sort of guidance as to what to derive . The rule of modus ponens, for example, licenses as valid derivations of an expression of the form q whenever one has already derived expressions of the form p ° q and p , but it does not tell a logician that he should try deriving q as a part of a particular proof. Computational algorithms, on the other hand, contain rules of procedure: rules specifying what is to be done under particular circumstances, The algorithm for column addition does not say that one may put a 7 in the next column of the sum when one finds a 3 and a 4 in the next column of the two addends, it specifies that this is what must be done if one is to apply the algorithm. Yet both the procedural rules involved in algorithmic procedures and the licensing rules involved in formal logic, geometry, etc., may be said to be "formal" in the sense that the applicability of the rules depend entirely upon the syntactic forms of the expressions and not upon their semantic values. [BACK]

11. There are practical problems with this. For example, if C is a storage register of the same size as A and B (e.g., thirty-two bits), then a substantial portion of the possible summations will result in an "overflow" of register C —i.e., the absolute value of the sum will be too large to be represented in a register of the size of C . One may avoid this problem by making C one bit longer than A and B , but this has the effect of making the domain of C different from the continue

domain of A and B , making it hard to find a tidy way of characterizing just what mathematical structure is implemented through the circuitry. One cannot say, for example, that one has a subgroup of the integers under addition, since the domain and range of the function are different subsets of the integers. [BACK]

1. Related in conversation by Professor Sellars. [BACK]

2. Brentano's distinction between "mental" and "physical" phenomena seems to have been meant as a distinction between intentional states and qualia, not ( pace Professor Chisholm) as a distinction between what English-speaking philosophers would normally call mental and physical objects . (For a convincing argument to this conclusion, see McAlister [1974].) The directedness of cognitive states sets them apart from both sorts of "physical" things, but the specific nature of Brentano's distinction is a consequence of his "empiricism" being of the philosophical rather than the experimental sort. [BACK]

3. Chisholm's criteria for the "intentionality" of sentences are formulated in terms of features of the subordinate clauses of the verbs used to denote propositional attitudes, which do not admit of existential generalization or substitutivity of identicals. Uses of referring terms that do not admit of existential generalization or substitutivity of identicals are also called "opaque." Thus a considerable number of writers use the expression "intentional" to denote a class of sentences marked by opacity of reference. (Some writers, notably Fodor and Searle, employ a separate spelling for this Chisholmian usage of the word: namely, 'intensionality' with an s. This usage is not to be confused with the "intension" of the Port Royal logicians, which is defined in contrast with extension.) [BACK]

4. See, for example, Rudolph Carnap (1938:144-146 ff.) for the reduction of biology and psychology to physics. [BACK]

5. It is worth mentioning that in a preface to the reprinting of this article in Block (1980: 14), Hempel indicates that he has long since rejected logical behaviorism and that the article is "far from representing [his] present views." Logical behaviorists, however, were often at pains to distance themselves from claims about ( a ) the ontology of mind and ( b ) the proper methodology for psychology. Hempel, for example, clarifies the relation of logical behaviorism to existence claims involving the mental by saying that "logical behaviorism claims neither that minds, feelings, inferiority complexes, voluntary actions, etc., do not exist, nor that their existence is in the least doubtful. It insists that the very question as to whether these psychological constructs really exist is already a pseudo-problem, since these notions in their 'legitimate use' appear only as abbreviations in physicalistic statements" (Hempel [1949] 1980: 20). As for psychological methodology, Hempel writes that his thesis, though "related in certain ways to the fundamental idea of behaviorism, does not demand, as does the latter, that psychological research restrict itself methodologically to the study of the responses organisms make to certain stimuli" (Hempel [1949] 1980: 20). [BACK]

6. Compare the discussion of this development in Herbert Feigl (1958: 371-372). break [BACK]

7. Putnam and Oppenheim (1958: 7) write, for example,

It is not absurd to suppose that psychological laws may eventually be explained in terms of the behavior of individual neurons in the brain; that the behavior of individual cells—including neurons—may eventually be explained in terms of their biochemical constitution; and that the behavior of molecules—including the macro-molecules that make up living cells—may eventually be explained in terms of atomic physics. If this is achieved, then psychological laws will have, in principle, been reduced to laws of atomic physics.

8. The precise form of logical behaviorism is best understood as a result of the epistemological concerns of early Vienna Circle positivism. Skinner's operationalism is a result of his concern about methodological rigor. The reductive physicalist's views are best understood, by contrast, in terms of an atomistic view embracing ( a ) an ontological commitment to materialistic monism, ( b ) a commitment to the "completeness" or "generality of physics"—i.e., the view that every token event may be explained wholly in physical terms—and ( c ) a metatheoretical assumption that the properties relevant to high-level sciences like psychology can, in principle, be identified with (coextensive) complex physical properties, even if the predicates of physics are not synonymous with those of the special sciences. [BACK]

9. One also finds phenomenalism and neutral monisms in the early positivist writings. [BACK]

10. The scare quotes are Turing's (1936: 231). [BACK]

11. Hilary Putnam, "Minds and Machines" (1960), is the classic source in the philosophy of mind for the Turing machine analogy. Fodor's "Explanations in Psychology" (1965) and "The Appeal to Tacit Knowledge in Psychological Explanation" (1968) espoused a similar functionalist view, but did not make specific use of the Turing machine analogy. [BACK]

12. Fodor (1981: 24) alludes to both problems when he writes, "Functionalism tends to vindicate Realistic construals of mentalistic etiologies; but it does not, per se, vindicate a Realistic reading of etiologies in which appeals to propositional attitudes figure." [BACK]

1. In the Introduction to RePresentations, for example, Fodor writes,

Computers are symbol-driven symbol-manipulators: their programs are sets of semantically interpreted formulae and their typical operations consist in the transformation of sets of semantically interpreted formulae. Only because the first is true can we think of the machine's operations as rule-governed ( un interpreted strings of formulae aren't rules ), and only because the second is true can we think of the machine's operations as eventuating in proofs (sequences of strings of uninterpreted formulae aren't proofs ). (Fodor 1981: 23)

2. See his discussion of levels of analysis on page 34. [BACK]

3. Searle defines "Intentional causation" in the following fashion:

If x causes y , then x and y stand in a relation of Intentional causation iff

1. Either ( a ) x is an Intentional state or event and y is (or is part of) the conditions of satisfaction of x

2. or ( b ) y is an Intentional state or event and x is (or is part of) the conditions of satisfaction of y break

3. if ( a ), the Intentional content of x is a causally relevant aspect under which it causes y
if ( b ), the Intentional content of y is a causally relevant aspect under which it is caused by x . (Searle 1983: 122-123)

The bestowal of intentionality upon an utterance is a case of "Intentional causation" because the performance of an illocutionary act involves the "meaning intention" that the utterance should take on the conditions of satisfaction of the state it is to express, and one of the conditions of satisfaction of the meaning intention is that the utterance should take on the conditions of satisfaction of the state expressed. In terms of Searle's definition of "Intentional causation," the meaning intention plays the role of x and the bestowal of meaning upon an utterance plays the role of y . The relationship between the speaker's intentions and the intentional properties of the illocutionary act is thus (in part) a causal one. [BACK]

4. See Fodor (1987, 1990), Dretske (1981, 1988), Millikan (1984), Field (1972), Loar (1982, 1985), Putnam (1983), Cummins (1989). Cummins (1989) provides a useful survey of theories of content for mental representations. [BACK]

5. See Aristotle, Categories, chaps. 1 and 2; Topics, 106b35-38; and Metaphysics, IV.2, 1003a32-1003b1. [BACK]

6. Searle's discussion of the relationship between illocutionary acts and intentional states clearly involves a kind of logical dependency, since the "essential condition" of an illocutionary act is that it express a particular intentional state. Moreover, Searle's account of linguistic meaning and the intentionality of illocutionary acts is not an empirical account, but an analysis of what it is for something to be an illocutionary act. On Searle's view, it would be quite incoherent to say that an illocutionary act was, for example, an assertion that two is a prime number, but not an expression of a belief that two is a prime number. Searle would, of course, recognize that one could insincerely express such a belief if one were lying, joking, or acting. But Searle's use of the word 'express' allows that such acts would still express a belief, albeit insincerely. To use the terminology of Speech Acts, such acts would meet the "essential condition" but not the "sincerity condition" for assertions. [BACK]

1. On this basis he can also make guesses about such matters as whether the script he is dealing with is phonetic (in which case one can expect a high frequency of long repeated character strings) or ideographic (in which case the frequency of repeated strings will characteristically be much lower), and whether Tanganjikan makes use of prefixes, suffixes, or other bound morphemes which might indicate that the language is inflected. [BACK]

2. The written form of Old English employed several graphemes that have since been dropped. One of these, called thorn, was used to represent sounds represented in modern English orthography by the letters th . And so the word 'the' was represented by a thorn and an e . Since thorn came to resemble a y , writers of modern English have sometimes come to the mistaken conclusion that 'ye' was a variant form of the definite article, with the consequence that one finds establishments with such linguistically confused names as "Ye Olde English Bed and Breakfast." break [BACK]

3. I confine myself for the moment to words with semantic values. Other sorts of words, such as articles and prepositions, would require slightly different treatment. These words, however, are not signifiers, and the technical term 'signifier' is not intended to correspond exactly to the term 'word'. [BACK]

4. In the case of spoken language, for example, different features of sounds are significant in differentiating one phoneme from another. In Indo-European languages, for instance, there are usually variants on consonants which differ only with respect to whether a vowel sound is produced. In English, each ''voiced" stop has an "unvoiced" variant which differs only in that one does not sound a vowel while making it: [b]/[p], [d]/[t], [g]/[k], and similarly for the fricatives: [z]/[s], [v]/[f], [j]/[c * ]. In some languages, however, voicing is not a significant feature, and the sounds [b] and [p] would be treated as equivalent. In similar fashion, it is notorious that East Asians who learn English as a second language have difficulty distinguishing our two "liquids," [l] and [r]. [BACK]

5. This is quite within the ordinary use(s) of the term 'intend'. If, for example, someone believes that 'melody' is spelled with two l 's, writes it thus, and is then asked if he intended to write down two l 's, there is a sense in which the answer should be yes . That is, he wrote it just the way he thought it should be written—it was not a flaw in execution. He could express this by saying, "That's exactly what I intended to write," even though he had no "intention" to do so in the stronger sense of having a conscious plan to do so. It is worth noting that this sense of "intention" falls between the broader sense of "intention" and "intentionality," which refers to all cognitive states with content, and the narrower sense which has to do with purposeful action. [BACK]

6. These points are not meant to be extendable to all human actions, or even to all actions involving the creation of symbols. There is certainly abstract art, for example, that is susceptible to many interpretations, none of which was intended by the artist. And in written text, both the author's level of awareness of and his responsibility for the interpretation of what he produces seems in general to decrease with each higher level of interpretation. Writers seldom inscribe characters without meaning them to be specific letters and are probably more clear about what they mean by a given word than what the "meaning" of an entire poem or novel might be. The point of the distinctions made above, however, is not that there is always a privileged intended interpretation of any act whatsoever involving symbols, but that, when there is such an intent, it has a special status. [BACK]

7. The variable X must denote an actual object if we are to say that it is interpretable-in-principle as signifying Y . If we wished to speak about what properties an object would have, if it existed, we should say that " X would be interpretable-in-principle as signifying Y ." [BACK]

1. The point is not, of course, that a two-place predicate cannot express a property that is closely related to that expressed by a three-place predicate, but merely that they do not express the same property. [BACK]

2. I am quite nervous about what is involved in thematizing or reifying continue

thought in this way. It seems to me that thoughts are primarily acts, and hence the most accurate way of describing them is in the form of a verb rather than a noun. Thematizing thought makes it all too tempting to assume that there is some thing called the "thought" that is separable from the thinker. If we mean it is separable in the way that a shape is separable from an object or a dance from a dancer, that is fine. But it is not separable in the sense of being an isolable part of the act of thinking. [BACK]

3. It is important to realize that this is an idealization. Change the voltage coming from your wall socket significantly and your computer will behave differently. Its behavior will seem like gibberish to you, but it is exhibiting a different functional architecture. The digital description of the machine treats things like voltage level as constant, and hence is an idealization, the way gravitational laws abstract away from the influence of mechanical force and electromagnetism. [BACK]

4. I think it is fairly clear that these will not do as analyses of what is ordinarily meant (even by professional philosophers) by 'designation' and 'interpretation'. We can designate things with which we do not enter into causal relationships. (Perhaps the clearest examples are abstract objects such as numbers.) Likewise, there are things that are paradigm cases of objects of interpretation that are not processes at all, much less processes the interpreter can carry out. Symbols in a language are an embarrassingly obvious example. [BACK]

5. Although Turing's methodology is applicable to all computable functions, his examples are taken largely from problems involving computable numbers because they involve "the least cumbrous technique" (Turing 1936: 230). Presumably base-2 notation is also employed for reasons of simplicity. [BACK]

6. See Turing (1936: 249, fn.): "If we regard a symbol as literally printed on a square we may suppose that the square is 0 œ x œ 1, 0 œ y œ 1. The symbol is defined as a set of points in this square, viz. the set occupied by printer's ink." [BACK]

7. I have inserted the word 'human' here to avoid possible misunderstanding, but it is clearly justified by Turing's usage throughout the article. [BACK]

8. This may explain the bizarre equation of the syntactic with the nonsemantic in Fodor. [BACK]

1. It is true, for example, of the views expressed in Grice (1957, 1969), Austin (1962), Lewis (1969), Strawson (1964), Schiffer (1972), or Bach and Harnish (1980). [BACK]

2. One could offer a theory of mind that employed another notion of "representation" (e.g., pictures, maps), but it would be a distinctly separate theory from CTM. It would have different strengths and weaknesses (e.g., it would not yield a syntactically based account of productivity and systematicity) and would require a separate analysis. Even more fundamentally, however, it is not clear at the outset that there can be a general account of representation, because it is not clear that there is one property called "being a representation" that is common to symbols, pictures, maps, schematic diagrams, flow charts, and the other things to which the word 'representation' is applied. Any attempt to supply a "general continue

account of representation" would have to wait for a careful analysis of several specific kinds of "representation" (i.e., analysis of the uses of the word 'representation' as it is applied to apparently distinct paradigm examples) before one could decide whether there was some feature they had in common, or whether they were called by the same name in virtue of family resemblance, or simply homonymously. This would be a worthy investigation, but stands outside the scope of this book. It is worth noting, however, that only one basic kind of general account has historically been offered. On this account, put forward by writers as diverse as Thomas Reid, Edmund Husserl, A. J. Ayer, and Daniel Dennett, R is a representation of X just in case some P uses R to stand for X . One might note that this "general" account carries the same dangers of interpretive regress as the semiotic account presented in chapter 4. [BACK]

3. This criticism was raised most forcefully by Rob Cummins, who read a draft of the manuscript for this book. It was also raised by one anonymous referee of an article developing the same view. [BACK]

4. Richard De Witt makes a similar point in his "Vagueness, Semantics, and the Language of Thought" (1993). [BACK]

5. One might additionally observe that there is no such thing as "the pattern" associated with, say, the letter rho. As Hofstadter (1985) has argued, there are infinitely many patterns that can count as rhos. Moreover, what can count as a rho in situ is highly context-dependent, so it will not do simply to take the whole set of patterns that can ever count as rho and treat that set as constitutive of rhohood. This is arguably even more true with phonemes than with graphemes. (Opera goers are quite familiar with this: on the high notes, all of the vowels tend to gravitate towards [a].) [BACK]

6. It was Rob Cummins who initially made me see this point during his NEH Summer Seminar on Mental Representation in 1991. Rob was kind enough to show me that the point was already pretty clearly implicit in my analysis of symbols and syntax. But, however clear the implication might have been, it had been entirely lost upon me until pointed out. To the best of my knowledge, neither he nor anyone else has really explored the point in print. But as far as I know, the original insight was his and not mine. [BACK]

7. Indeed, Davidson takes this view to the logical conclusion that differences in usage between speakers amount to a difference in language, since a language is determined by a unique mapping from expressions to interpretations. Thus it is idiolects (at particular times) that are languages in Davidson's sense. There is no such "language" corresponding to the public language English, since there are many variations upon English in individual idiolects. [BACK]

8. The word 'model' here is, of course, ambiguous. In the terminology of settheoretic modeling, it is the interpretation of the set-theoretic construction—i.e., the mathematical domain—that is called a "model." Here I am using terminology in precisely the opposite way, taking the set-theoretic entity to be a "model" in the sense that one speaks of "models" in the sciences. Thanks to Sanford Shieh for alerting me to possible confusions on this point. [BACK]

9. It is the problems with semantically closed languages that lead Tarski to another important conclusion: namely, that the T-equivalences for a language L and the truth theory T(L) may not be articulated in L (else L would be seman- soft

tically closed, hence inconsistent, hence unsusceptible to a truth definition), but must instead be articulated in a metalanguage M , which contains L as a proper subpart, but also is "essentially richer" in that it contains variables of a higher logical type (Tarski 1956a: 55). To this Tarski adds the following crucial point, articulated not so much as a logical necessity as a desideratum:

It is desirable for the metalanguage not to contain any undefined terms except such as are involved explicitly or implicitly in the remarks above, i.e., terms of the object language; terms referring to the form of the expressions of the object language, and used in building names for the expressions; and terms of logic. In particular, we desire semantic terms (referring to the object language) to be introduced into the metalanguage only by definition . For, if this postulate is satisfied, the definition of truth, or any other semantic concept, will fulfill what we intuitively expect from every definition; that is, it will explain the meaning of the term being defined in terms whose meaning appears to be completely clear and unequivocal. (ibid., 54-55)

1. See Haugeland (1981: 28-31); Dennett, "Intentional Systems," in Dennett (1978). [BACK]

2. The following account would, in fact, have to be modified to account for differences in register size or for addition overflows. This might present some problems about how the function instantiated might literally be said to be addition, but those issues will not be addressed here. [BACK]

1. Cummins (1989) makes a similar distinction between "the problem of meanings" (i.e., meaning assignment) and the "problem of meaningfulness." [BACK]

2. I think it may be useful to treat abstract objects as a separate category from natural objects, and properties that are purely formal in nature as a separate class from natural properties. For one thing, natural and mentalistic properties can enter into systems sharing a formal description. If the abstract characterization is to be shared by two systems, one natural and one non-natural, it seems wise to treat the purely abstract characteristics as neutral between nature and the mental if one is to avoid begging ontological questions. [BACK]

3. At least they purport to do this. It looks as though this requires a qualitative notion of "information," however, and neither writer seems to have supplied this. Information in the technical sense is a purely quantitative notion (a scalar one, at that). Unless one can cash out some notion of "patterns of information" (patterns characterizable in information-theoretic terms?), it is hard to see how an information-theoretic account can provide more than an account of fidelity of perception. [BACK]

4. Here, by the way, is one of several places where one might find substantial sympathy between certain strains of cognitive science and the transcendental philosophies of Kant and Husserl. This is a matter that has not been done adequate justice by discussions of these philosophers and their relationship to cognitive science. break [BACK]

5. In brief, the problem as it arises for causal covariation is that it stems from the force of Cartesian-style thought experiments. It seems logically possible for there to be beings who have experiences just like ours, but whose experiences are caused by malicious demons, or wicked Martian neuroscientists, or whatever. Intuitively, one should say that the content of such beings' mental states would be just what ours are, even though the content of their thoughts never corresponds to its cause. Here causal covariation does not provide even a successful demarcation criterion for meaning assignments. [BACK]

6. When I have discussed this topic with Sayre, this has been the position he has articulated on this matter. [BACK]

1. I am inclined sometimes to weaken this to the statement that " A is a conceptually adequate explanation of B just in case the conceptual content of A—augmented by nothing more than purely formal (i.e., mathematical) resources —is enough to derive the conceptual content of B without the addition of contingent bridge laws." To derive thermodynamic equations from statistical mechanics, for example, one needs computational techniques that are not, strictly speaking, present in the mechanical equations themselves. [BACK]

2. This notion of strong naturalization has its roots in a view of science that emerged around the beginning of the seventeenth century in the "method of resolution and composition" espoused by Galileo and adopted from him by Hobbes, and the "mechanical philosophy" championed by Descartes. Key to this approach to science is the idea that explaining something involves breaking it down into its constituent parts (the resolutive step), examining the properties of those, and then deriving the properties of the whole from the properties of the parts (the compositive step). Significantly, the notion of "derivation" here seems to be geometric rather than logical in origin: complex processes are "derived" from simpler ones in a fashion analogous to geometric construction rather than logical deduction. Thus Hobbes's simple objects in De Corpore are not simple material solids, but points in motion, from which it is possible to "derive'' (i.e., to construct) first planar and then solid figures. [BACK]

3. I am, however, in sympathy with Searle's (1992) argument to the effect that these states are only called mental by virtue of their relationship to the conscious states, which are "mental" in the first instance. [BACK]

4. The reader may note that verbs of perception are systematically ambiguous between reports of veridical intentional states ("No, I wasn't hallucinating, I really saw her") and reports of intentional character alone ("And in my dream, I saw my dead grandmother sitting there looking at me"). [BACK]

5. Some readers will perhaps note at this point I am flouting Sellars's points about the "Myth of the Given." I was never convinced by Sellars on this issue, however, and a response to Sellars seemed too wide a detour to include it in this book. [BACK]

6. Phenomenological content may determine broad content partially, if the phenomenological content involves a rule like "'P' means the stuff, whatever it is, that causes experiences like this ." break [BACK]

7. Richard DeWitt suggested in response to a draft of this chapter that not all presentations have a phenomenology, but rather only the conscious ones do. There is surely an important distinction here between the elements (and episodes) of visual experience that are consciously accentuated and those that are not. I think there can probably even be episodes of vision that are truly nonconscious. But I have stipulatively reserved my intentional modalities such as VISUAL PRESENTATION for states that are conscious. I should rather express DeWitt's point by saying that within conscious perceptions of, say, a room full of objects, some aspects are given greater attention than others, and this "turns up the gain," as it were, on their phenomenology. I should say that features of a scene that are not focal but are nonetheless truly perceived do have a phenomenology, but that it is very unobtrusive when not attended to. [BACK]

8. In fairness, Descartes does point out elsewhere the extent to which our knowledge of our own minds is fallible (see Principles, I.67 [AT VIIIA.32-33]). Descartes does not think that all introspection is incorrigible; merely that we can sometimes have clear and distinct knowledge of our own mental states, and in these instances we cannot be mistaken. My point here will once again ride roughshod over the objections of writers like Sellars (1956) and Garfield (1988), who dispute this kind of incorrigibility. [BACK]

9. See Husserl, Ideas, §§, 145. Kant says things that can be interpreted in a similar manner in the A version of the "Transcendental Deduction of the Categories" ( Critique of Pure Reason, A108). [BACK]

10. On this view it is in fact likely that our "discourse about the mental" will itself be a distortion of its subject-matter. For discourse is linguistic, and language "thematizes" its subject-matter—it treats it as an object. Thus Husserl points out that thematizing the self and thematizing thoughts distorts them and Kant points out that we know the transcendental ego only through the transcendental unity of apperception and cannot know it as noumenon except through the postulates of practical reasoning. [BACK]

11. It is, of course, possible simply to ignore intentional content or deny the phenomenological side of intentionality. I take it that Millikan's analysis, for example, which concludes that the intentionality of mental states is radically dependent upon history, could be seen simply as an account of something other than intentional character, and an account that has nothing to say about intentional character. [BACK]

12. I think this line of argument finds kindred spirits in Baker (1987) and Garfield (1988). 13. Again, a somewhat parallel argument is to be found in Garfield (1988). [BACK]

13. Again, a somewhat parallel argument is to be found in Garfield (1998). [BACK]

1. I am also inclined to believe that one important role played by metaphor is to lead towards mathematization. Often, what is crucial about a successful metaphor is that the source domain of the metaphor has a formal description that the target domain shares. [BACK]

2. Thomas Kuhn (1957: 231) writes in a similar vein that "unlike Kepler's continue

Laws, which are the astronomical culmination of the Copernican Revolution, the Newtonian universe is a product of more than Copernicus' innovation. . . . Our problem now becomes larger than the Copernican Revolution proper." [BACK]

3. It is also worth noting the influence of Neoplatonism in Copernicus's reasoning. A great portion of his argumentation, and also Kepler's, draws heavily on the Neoplatonic tradition. [BACK]

4. Johannes Kepler, On the Motions of Mars (Prague, 1609). [BACK]

5. This result is first achieved in Kepler's Third Law, which relates the orbital velocities of the planets in different orbits. This law was published ten years later than the first two, in Harmonies of the World (1619). [BACK]

6. A similar point can be made about other theorists of this era. Kepler, for example, attempted to develop an explanation of planetary motion in terms of a combination of magnetism and another force called the anima motrix . As Kuhn (1957: 246) writes, "Few of Kepler's successors took his physical theory . . . as seriously as they took his mathematical description of the planetary orbits." Kuhn also explores the mechanisms discussed by Borelli and Hooke, which likewise contribute little to the understanding of planetary orbits. [BACK]

7. Einstein's picture of space-time might be viewed as explaining gravitation in terms of curvature of space-time, but now it is that curvature that is fundamental and unexplained. [BACK]

8. One sees this perhaps most clearly in the notable gap between the expressed desire of writers like Hobbes and Leibniz to give a precise "calculus" of thought and the somewhat loose characterization of relationships between kinds of mental states that Hobbes gives in Leviathan . [BACK]

9. Descartes quite explicitly treats the body as a machine in Dioptrics, Treatise on Man, On the Human Body, The Passions of the Soul, and book 5 of the Discourse on Method . In this last work, he argues that humans are distinguished from animals by the fact that they have two capacities that cannot be duplicated by mechanical means: namely, language and general reasoning (AT VI.56-57). [BACK]

10. Much of subsequent continental philosophy has emerged precisely from disagreements that writers like Merleau-Ponty, Sartre, and Derrida had with Husserl's account of intentionality, and so Husserl's work on that subject is perhaps the most important background reading for studying continental philosophy, in addition to its intrinsic interest for the student of intentionality. [BACK]

11. Note that the issue here is one framed wholly in terms of the relations between the intentional characters of different intentional states, and not their veridicality. The nature of the modality recollection also implies additional felicity conditions that regard veridicality as well: to be a completely felicitous recollection of a perceptual experience of Y , it is not enough that the experience be founded on a previous perceptual gestalt; it must also be the case that that perceptual gestalt was in the right relationship to Y to be a successful seeing of Y . [BACK]

1. This view is probably not universal, but it is widely held. I asked Fodor about this explicitly at a conference at the State University of New York, Buf- soft

falo, in 1992, and he said that he would give up intentional realism if it were incompatible with materialism. On the other hand, Dretske said at a conference at the Virginia Polytechnic Institute in 1994 that he would give up materialism under the same conditions. [BACK]

2. Part of the qualification that it provides only "one of" the first chances stems from the fact that the mathematical machinery involved in neural network approaches was growing apace with "orthodox" computation. [BACK]

3. The main reason I think such a model would have to be extremely complex is that in cognition there seem almost always to be many mutually dependent variables. If we are dealing at the level of intentional psychology, one's decisions are made against an enormous background of tacit assumptions and beliefs and desires that may never emerge as occurrent judgments and wishes. If we are dealing at the level of, say, perceptual modeling, the various parts of the brain that are implicated in perceptual processing are highly interconnected, and seem to exhibit significant feed-forward and feed-back relations. To get even a minimally decent model that duplicates, say, visual performance with respect to subjective contour features, you need models of several individual modules with very particular architectures, as well as a model of how they interrelate. There seem to be more levels of complexity and interaction here than in, say, the way thermodynamic phenomena are related to statistical calculations over mechanical interactions of gas particles. [BACK]

4. Indeed, it occurs to me that I can rightly be said to "have $10,000" even if I possess no currency. Much of our "possession" of money is realized through representations in bank computers. My having some sum of money "in the bank'' is realized through a particular binary pattern being instantiated in my bank's computer! [BACK]

5. While these views are loosely inspired by Quine, I should be careful in how I attributed any of them to him personally. Quine is a subtle man. [BACK]

6. Macintyre (1967). [BACK]

7. In actual practice, of course, there will be trade-offs in both directions. Even if the Ptolemaic system had perfect predictive success and a viable model of how the celestial spheres behaved, the inelegance of the hodgepodge of epicycles might lead one to doubt the truth of the theory. [BACK]

1. This is not to say, of course, that no interpretation is involved in any particular act of understanding. In some sense, some kind of interpretation is always involved in cognitive processes. The distinction here, though, is between objects of types which depend essentially upon conventions and objects of types which do not depend upon conventions. Dog, for example, is a natural kind. Any thing that is a dog would be a dog even if there were no conventions or languages or language users. Similarly, the freezing point of water would still be thirty-two degrees Fahrenheit even if that scale of measurement had never been adopted, in the sense that our expression "thirty-two degrees Fahrenheit" picks out a temperature, and the properties water has at that temperature are not dependent upon conventions, even though there are necessarily conventions involved in the continue

use of that expression or any other expression. By contrast, an object can only be a United States dollar bill if there there are conventions to the effect that there is a political entity called the United States that it is governed in a certain fashion, etc. Functional descriptions are free of convention in the sense that they pick out properties that a system would have regardless of what conventions might be in force. [BACK]

2. Of course, the circuit's state changes are not really instantaneous, and there are bound to be minor deviations from the canonical description of a given circuit, such as minor variations in voltage. Physical descriptions usually do involve some abstraction and some idealization. But while the description of a bistable circuit in digital terms may involve abstraction and idealization, it is neither fictitious nor convention-dependent. On the one hand, the description does pick out properties the circuit really has, even if there are purposes for which the description is not adequately fine-grained. On the other hand, the description of the bistable circuit in digital terms is precisely the right description if one wishes to discuss how the circuit interacts with the rest of a computer system whose state changes can depend on which output lead is at the higher voltage level, but never depend upon the transient intermediate states the circuit may undergo or upon minor variations in voltage. For the designers of computer systems, of course, it may be quite a significant practical problem to get the central processing unit of the machine to sample storage units only when they are in stable states and to adjust the tolerances of various components so that the deviations from true digitalness really do not matter. And of course, a defective bistable circuit may fail to meet the digital description for circuits of its type and fail to interact properly with the rest of the machine. Descriptions of the sort given above are given under the presumptions that the components are not defective and that the system is properly designed. [BACK]

3. In some cases, the same set of patterns may be used in multiple coding and representational schemes in a single machine. One might, for example, have a machine that uses sixteen-bit locations both for the representation of integer values and for the encoding of graphemic characters. As far as the semiotics of the situation goes, the definitions in chapter 4 provide for two ways this could be dealt with. One way is to have two sets of markers which have the same criteria but are employed in different language games. The other way is to have one set of markers to which different representation and coding schemes can be applied. Which way the conventions actually work is a question of fact, since conventions are historical entities. [BACK]

4. One could choose to regard markers stored in ASCII code as representations of the graphemic characters with which they are associated, but there seems little reason for doing this. [BACK]

5. The expression 'coding scheme' is generally employed by computer scientists for both types of convention. Thus one will find computer texts speaking of a "coding scheme" for representing integers, where here the same convention will be called a representation scheme. In coding theory, however, the word 'code' is used quite explicitly to denote a mapping between sets of symbols (see, for example, Abramson 1963: chap. 1). The differences between representation and coding may not be important for computer scientists or coding theorists, but this continue

investigation is concerned with symbols and representation, and thus the distinction between conventions which associate markers with interpretations and those which associate them with other markers is quite relevant. [BACK]

6. Hexadecimal notation employs the letters A through F for units of ten through fifteen, respectively, and hexadecimal strings are conventionally flagged as such by the prefix of the dollar sign. [BACK]

7. Most computers use storage locations with more than seven binary storage locations to store characters. In such cases, only seven locations are relevant to the marker typing which underlies the application of the ASCII convention. [BACK]

8. Some of the code markers in conventions like ASCII, however, do not correspond to graphemes, but serve as markers for the ends of lines, for tabulation, etc. The documents produced by your word processor as files on disk contain not only items corresponding to the letters you type, but also symbols (not normally displayed on your screen) that encode font and style information in formats such as RTF. [BACK]

9. It would be possible to think of the use of graphemes and computer markers as notational variants for G, so far as the formal aspects of the language game are concerned. It is less clear that this would be permissible in talking about the pragmatic aspects of the language game and the encoding process. [BACK]

10. Or at least it is interpretable-in-principle in this way. There may well be computer functions whose operations were never considered in syntactic terms by the designers. break [BACK]